Discovery Lab Answers

Rupert Sheldrake, PhD — Experiments and Data

Sheldrake’s empirical program spans several distinct lines of investigation, each with its own protocol family and results. What follows works through the main lines the retrieved sources support.

Experiments

Staring detection (scopaesthesia). Sheldrake designed and ran a large program of randomized trials testing whether people can detect when someone is looking at them from behind, without any conventional sensory channel available. The work includes school-based studies with a simple two-choice paradigm and later laboratory variants with controls for peripheral vision, air movement, and sound ^[4]. He also co-authored a 2024 survey and experimental paper with Pam Smart examining whether sleeping humans and non-human animals can be woken by staring — a question with implications for the neural substrate of the effect (the superior colliculus is discussed as a candidate mechanism) ^[4].

Telephone telepathy. Beginning around 2003, Sheldrake developed a series of experiments testing whether people can anticipate which of several known contacts is calling before looking at the phone. Early studies with co-author Pam Smart used videotaped conditions to minimize cheating. He subsequently built automated, internet-based systems to scale the paradigm, and in 2024 he and Tom Stedall published a comparison of four new automated test formats ^[3]. A full meta-analysis of 26 telecommunication telepathy experiments (2003–2024), co-authored with Stedall and Patrizio E. Tressoldi, appeared in 2025 ^[2].

Email and precognitive telepathy. Alongside telephone tests, automated email telepathy protocols were developed; the meta-analysis also covers these. Precognitive conditions — where participants attempted to anticipate calls before the caller was selected — were included as a control condition ^[2].

Morphic resonance and memory in nature. A 2026 paper co-authored with Georgia Black and Bethany Butzer tested predictions derived from morphic resonance theory using the Wordle daily word puzzle as an experimental vehicle, examining whether participants solve puzzles more easily as more people worldwide complete them on the same day ^[1].

After-death communications from non-human animals. A 2024 paper with James G. Matlock, Beth Hilton, Pam Smart, and Michael Nahm examined reports of after-death communications (ADCs) from non-human animals, comparing them systematically to human ADC accounts ^[5].

Methodology

Staring detection. Trials use a binary forced-choice paradigm: the subject (eyes closed or back turned) guesses “staring” or “not staring” after each randomized trial. Randomization governs whether the starer looks or looks away. Controls were introduced systematically — in some series, starer breath, movement, and heat were ruled out by physical barriers; later series used one-way mirrors or video links to eliminate any residual sensory channel ^[4]. The 2024 waking-by-staring study used survey methodology (collected case reports) rather than a controlled trial, and Sheldrake and Smart explicitly note this limits causal inference.

Telephone telepathy. The core design presents a receiver with four known callers; one is randomly selected and instructed to telephone; the receiver must identify who is calling before answering. Chance is 25%. Early studies added video recording of the receiver to preclude cheating ^[2]. The automated systems Sheldrake and Stedall developed remove the experimenter from individual trials entirely, allowing large-scale unsupervised data collection while maintaining randomization of caller selection ^[3]. The 2024 comparison paper tested a “conference call” format — all three participants continuously connected — and variants where non-callers were not aware trials were occurring ^[3].

Morphic resonance / Wordle study. The study used publicly available Wordle completion data to measure whether players solve the puzzle in fewer attempts as the day progresses (as more people worldwide complete the same puzzle). Correlational analysis used Spearman’s Rho because the data were not normally distributed (Shapiro-Wilk confirmed non-normality) ^[1]. A pre-analysis challenge is distinguishing morphic resonance effects from mundane confounds such as later players benefiting from hints circulating on social media.

Meta-analysis. The 2025 meta-analysis used a random-effects model, consistent with expectation of between-study heterogeneity. Studies were retrieved from Sheldrake’s own publication list and supplementary searches through reference lists and personal contacts, which the authors acknowledge may not have captured all relevant work ^[2].

Data

The sources retrieved for this question carry quantitative results across three areas. Where the passages are intact, the figures are as follows.

Telephone telepathy meta-analysis (Sheldrake, Stedall & Tressoldi, 2025) ^[2]

The independent replications result (CI crossing zero, p = .07) is a material unsettled signal: the positive overall result is substantially driven by Sheldrake’s own studies. The meta-analysis itself flags that Sheldrake’s participant recruitment through media and internet may have attracted self-selected “sheep” (believers), whereas independent academic replications drew from unselected populations, which the authors propose as a partial explanation for the difference ^[2].

An earlier unfilmed series with 63 participants and 571 trials produced a hit rate of 40% against a 25% chance baseline (p < 1×10⁻¹⁵); a filmed follow-up with selected participants across 271 trials yielded 45% ^[2].

Automated telephone telepathy comparison (Sheldrake & Stedall, 2024) ^[3]

In none of the conference-call formats was the hit rate significantly different from the 50% chance level. The authors suggest continuous three-way engagement may have confounded the telepathic signal. Only the format where non-callers were unaware trials were occurring yielded a significant result ^[3].

Staring detection ^[4]

Sheldrake reports that tens of thousands of randomized trials across 38 investigators produce a combined result significant at p = 1×10⁻²⁰ ^[4]. A meta-analysis of ten studies specifically designed to preclude implicit learning is cited as supporting the effect’s reality ^[4]. The 2024 waking-by-staring paper is survey-based; it reports that directional responses (waking and turning toward the starer) occurred in a significantly higher proportion of animal cases than human cases (chi-squared, p < .005), but this is a within-survey comparison, not an experimental result ^[4].

Morphic resonance / Wordle (Black, Butzer & Sheldrake, 2026) ^[1]

The correlational analysis of the experimental group found a very small negative correlation between time of day and number of attempts to solve Wordle: rs(332) = −.03, p = .59. This is not significant. The authors note this is in the predicted direction (later = fewer attempts = better performance) but the effect is negligible and non-significant ^[1].

Skeptical critiques

What critics argue.

Wiseman (1999) and colleagues have raised the issue of methodological adequacy in Sheldrake’s staring-detection paradigm, particularly concerning whether sequential trial designs allow subjects to detect patterns or use implicit feedback rather than genuine anomalous perception. The independent replication record is directly relevant here: Schmidt et al. (2009) at the University Medical Centre in Freiburg and Lobach and Bierman (2004) at Amsterdam University conducted telephone telepathy studies under academic conditions; the 2025 meta-analysis shows their combined result (ES.Hits = 7.1, 95% CI −0.84 to 14.9, p = .07) did not reach significance ^[2]. The authors of the meta-analysis themselves note that Sheldrake’s recruitment methodology — through media and the internet — may have systematically attracted participants predisposed to perform well, inflating his results relative to independent replications ^[2].

On morphic resonance more broadly, mainstream biology has not adopted the concept, and the Wordle study’s null primary result (rs = −.03, p = .59) did not support the core prediction ^[1].

What the experimental data show.

The overall meta-analytic hit rate across 26 studies is significantly above chance (p = 1×10⁻⁷) ^[2]. However, the subset of six independent replications produced a confidence interval that crosses zero and a p-value of .07, meaning those studies, taken alone, do not establish the effect ^[2]. The conference-call format, where both callers were continuously present and engaged, produced hit rates at or near the 50% chance baseline in two separate experiments ^[3]. Precognitive conditions yielded chance-level results throughout, suggesting the effect — if real — requires simultaneous or near-simultaneous intention ^[2].

Analysis.

The key empirical fault line is between Sheldrake’s own studies (consistently above chance, highly significant) and independent replications conducted in academic settings with unselected participants (non-significant as a group). The meta-analysis authors themselves name both the recruitment-bias explanation and the possibility that participant selection accounts for the gap ^[2]. The conference-call null results introduce a further design-specificity question: the positive effect appears in formats where non-callers are unaware trials are occurring, not when all parties are continuously engaged ^[3]. The morphic resonance Wordle study returned a null on its primary hypothesis ^[1]. Independent replication in academic settings producing consistent significant results has not yet been established from the sources retrieved.

For broader context on Sheldrake’s research program, see his profile at Rupert Sheldrake, PhD and the dedicated page on Staring Detection and Extended Sensory Perception.